UDF's in redshift : Possible to reference a udf within another

Question

Is possible to nest UDF's within each other ?

Following is a code for computing confidence intervals in A/B tests - Ofcourse, I could write a huge function that does all-in-one, but wondering a better way to achieve this goal ?

set search_path to public;
create function cumnormdist(x float)
  returns float
IMMUTABLE AS $$
  import math
  b1 = 0.319381530
  b2 = -0.356563782
  b3 = 1.781477937
  b4 = -1.821255978
  b5 = 1.330274429
  p = 0.2316419
  c = 0.39894228
  h=math.exp(-x * x / 2.0)
  if(x >= 0.0) :
    t = 1.0 / ( 1.0 + p * x )
    return (1.0 - c * h * t *( t *( t * ( t * ( t * b5 + b4 ) + b3 ) + b2 ) + b1 ))
  else :
    t = 1.0 / ( 1.0 - p * x );
    return ( c * h * t *( t *( t * ( t * ( t * b5 + b4 ) + b3 ) + b2 ) + b1 ))
$$ language plpythonu;


set search_path to public;
create or replace function conversion(experience_total float,experience_conversions float)
  returns float
IMMUTABLE AS $$
  return experience_conversions*1.0/experience_total
$$ language plpythonu;


create or replace function zscore(total_c float,conversions_c float,total_t float,conversions_t float )
  returns float
IMMUTABLE AS $$
  import math
  z = conversion(total_t,conversions_t )-conversion(total_c,conversions_c) # Difference in means
  s =(conversion(total_t,conversions_t)*(1-conversion(total_t,conversions_t)))/total_t+(conversion(total_c,conversions_c)*(1-conversion(total_c,conversions_c)))/total_c
  return float(z)/float(math.sqrt(s))
$$ language plpythonu;

create or replace function confidence(total_c float,conversions_c float,total_t float,conversions_t float )
  returns float
IMMUTABLE AS $$
  import math
  return **(1-float(cumnormdist(zscore(total_c float,conversions_c float,total_t float,conversions_t float )),4))*100.00**
$$ language plpythonu;

The individual calls work fine, eg : select cumnormdist (-3.1641397476); If I insert them in the function definition, they don't, for example zscore that calls conversion function.

ERROR:  NameError: global name 'zscore' is not defined. Please look at svl_udf_log for more information
DETAIL:  
  -----------------------------------------------
  error:  NameError: global name 'zscore' is not defined. Please look at svl_udf_log for more information
  code:      10000
  context:   UDF
  query:     0
  location:  udf_client.cpp:298
  process:   padbmaster [pid=3585]
  -----------------------------------------------

If I could nest functions inside each other,(instead of having UDF's as above that are finally nested) that would be a reasonable status-quo.

End goal : Publish these computations in Tableau.

Answer 1

Here's how I solved it. UDF's cannot cross-reference the contents of another UDF, so you can create a custom library, upload it to AWS using CREATE library.

More here